Work

Equilibria, Convergence, and Basin Geometry in Adaptive Weighted Networks

Year: 2026
Type: preprint
Abstract: We study a finite adaptive weighted-network model in which node states follow symmetric Laplacian consensus dynamics and candidate-edge weights adapt to squared state differences with logistic saturation and linear decay. On arbitrary undirected candidate graphs, we characterize equilibria by their positive-edge support, derive the complete transverse spectrum of an arbitrary support equilibrium, and establish convergence of every trajectory to a single equilibrium using a common Lyapunov function, a sorted-sum argument, and scalar comparison. We then develop topology-aware outcome certificates: a protected-backbone cut criterion guarantees exponential consensus, an initial-data cut bound can rule out consensus, and under connected positive-threshold support consensus is equivalent to eventual success of a finite-time protected-spanning test. On an exactly invariant two-community family, the reduced orbit equation is linear after an integrating factor. This yields an explicit integral separatrix that completely classifies every positive cross-edge seed into fusion, strictly stable separation, or a critical boundary orbit, together with the separated endpoint. Near the separatrix we obtain a square-root law for the separated endpoint, algebraic critical convergence, and explicit ε⁻¹ᐟ² time scales on the fusion side. Targeted numerical integrations agree with the predicted equilibria, thresholds, quotient partition, endpoint formulas, and critical time scales within the stated finite-tolerance protocols.
Language: English
Open Access status: green
OpenAlex ID: W7236712904
Cites:
Cited by:
Related to: 10